The rise of AI in enterprises is reshaping cloud strategy. Sensitive data, strict compliance mandates, and the need for predictable performance are pushing organizations to reconsider how “cloud” should work. And they’re finding that modern private clouds, not public hyperscale platforms, offer the right balance of control and agility for running next-gen workloads.
According to the Private Cloud Outlook 2025 study of 1,800 IT leaders, private cloud is re-emerging as a strategic platform for innovation, supporting everything from monolithic VMs to containers to AI/ML workloads.
Key findings from the report reinforce this evolution:
- Over half (53%) of enterprises say deploying new workloads to private cloud is among their top priorities over the next three years.
- 84% already run both traditional and cloud-native apps in private environments.
- 69% are considering repatriating workloads from public to private cloud, and more than one-third have already done so.
Why private cloud for AI?
Privacy, security, and compliance
Generative AI (GenAI) turns your proprietary data into a competitive advantage, so where that data resides matters. Forty-nine percent of enterprises cite privacy, regulatory concerns, and security among the top challenges for adopting GenAI. Ninety-two percent say they trust private cloud for security and compliance, which is a top reason for repatriation. In turn, this underscores the importance of private AI when it comes to sharing private data with third parties and the associated new risk potential for organizations.
Regulatory pressure is increasing worldwide. The EU AI Act’s obligations began taking effect in 2024 and will follow a staggered timeline until fully applicable in August 2026. These requirements make data locality, access control, and traceability non-negotiable for many sectors.
Private Cloud and AI unlock choice and performance
As AI rapidly expands, it’s crucial for enterprises to select multi-vendor accelerators and large language models (LLMs) that align with their specific GenAI needs, industry requirements, and organizational goals. Making a strategic and flexible choice helps ensure the selected LLMs are effective for today’s use cases and adaptable to future demands.
AI and GenAI require significant infrastructural resources. Fine-tuning, customizing, deploying, and querying them can be intensive. Scaling up can be challenging without adequate resources. Efficiently allocating and balancing specialized hardware, such as GPUs, across the organization is critical to enable low latency and fast response times. In addition to GPUs, network and data I/O performance are equally critical to successful AI deployments.
Cost visibility for sustained compute
Public cloud GPU pricing, egress fees, and usage variance (i.e., token-based billing) make cost forecasting difficult. In contrast, private environments deliver long-term cost visibility and more predictable TCO.
According to InfoWorld, “It’s crucial to account for the cost of data gravity and migrations. Moving petabytes between cloud and on-prem environments can be both technically complex and financially overwhelming.” Instead, it makes more sense to bring the compute to where the data is.
A private cloud platform for private AI
The private cloud of today looks nothing like the private cloud of ten years ago. VMware Cloud Foundation (VCF) is proof of that transformation. It replaces rigid, siloed infrastructure with a unified platform that’s ready for any workload.
VCF brings together compute, storage, networking, security, and management in a single, integrated platform. It gives enterprises a true cloud operating model for private infrastructure—one that scales easily, enforces compliance, and supports everything from traditional enterprise apps to modern, data-intensive workloads.
That same flexibility and governance make VCF the natural foundation for AI-ready infrastructure.
Private AI, Built In
At VMware Explore 2025, Broadcom announced that VCF Private AI Services will become a standard component included with VCF. This integration will transform VCF into an AI-ready private cloud, where enterprises can run, govern, and scale AI workloads alongside every other workload they manage. (Read the press release.)
Here are some of the top capabilities enabled through VCF Private AI Services:
Model Store
Enterprises lack the proper governance measures for downloading and deploying LLMs from ISVs like Hugging Face or other third-party sources. With the introduction of Model Store capability, ML Ops teams and data scientists can now curate and provide secure LLMs with integrated role-based access control (RBAC). This can enable governance and security for the environment and the privacy of enterprise data and IP.
Model Runtime
The Model Runtime service enables data scientists to create and manage model endpoints for their applications. This simplifies model usage and enhances the scalability of LLMs by abstracting away the complexity of individual model instances and enabling deployment on servers that can handle multiple requests concurrently, rather than running the model locally for each request—which can be resource-intensive.
Agent Builder Service
AI agents are autonomous or semi-autonomous software entities, increasingly being integrated into GenAI applications, to perceive, make decisions, take actions and achieve goals in their digital or physical environments. The Agent Builder Service allows for GenAI application developers to build AI agents by using resources from the Model Store, Model Runtime, and Data Indexing and Retrieval Service.
API Gateway
Organizations face challenges integrating LLMs into applications, from security risks due to unprotected endpoints and frequent API changes to scalability issues and inconsistent APIs across providers. API Gateway addresses these challenges by providing a secure, stable, and scalable interface for accessing LLM endpoints, enabling strong authentication and authorization, and activating seamless integration with consistent performance.
Data Indexing and Retrieval Service
This service allows enterprises to chunk and index private data sources (e.g., PDFs, CSVs, PPTs, Microsoft Office docs, internal web or wiki pages) and vectorize the data. This vectorized data is made available through knowledge bases, which provides logical separation of indexed data so that IT administrators can enforce access controls, ensuring that interactions with models are limited to only the data a particular user can access. As data changes, these knowledge bases can be updated on a schedule or on demand as needed, ensuring that GenAI applications can access the latest data. This capability reduces deployment time, simplifies data preparation, and improves GenAI output quality for data scientists and ML Ops teams.
Multi-Tenancy
With VCF 9.0's multi-tenancy capability, cloud service providers and enterprises can enable secure and private environments for tenants on the same infrastructure and achieve high efficiency, scalability, and lower TCO. Private AI deployments can be done independently for each tenant.
VMware Private AI Ecosystem
Broadcom’s broad AI ecosystem includes industry leaders — from technology partners like NVIDIA, AMD, and Intel to OEM partners including Dell, HPE, Lenovo, and Supermicro. Broadcom also partners with global system integrators such as SHI, WWT, HCL, Wipro, Kyndryl, and more, to deliver enterprise-grade solutions at scale. Broadcom further collaborates with cutting-edge artificial intelligence partners such as Hugging Face, Domino Data Lab, and more to accelerate innovation and unlock the full potential of AI for our customers.
Bottom Line
We all know that AI workloads are here to stay. The Private Cloud Outlook 2025 shows that enterprise organizations want the benefits of GenAI; 98% of them are somewhere on the adoption curve, and 77% are already running pilots or live deployments.
With VMware Cloud Foundation shipping AI-native capabilities as part of the platform, enterprises will be able to scale AI services wherever their data lives.
“AI-native means embedding crucial AI capabilities directly into the core infrastructure,” says Paul Turner, Vice President of VMware Cloud Foundation Division at Broadcom. Doing so enables customers “to offer AI as a governed and secure service within their private cloud, from fine-tuning to inference.”
Learn more about how VMware Private AI can help you deploy AI across your organization with confidence. For more information about how other enterprise organizations are evolving their cloud strategies, read the Private Cloud Outlook 2025.